Monday, 22 July 2013

What is there to be afraid of?


What is worthy of our fear? Is Death? Surely not. He who lives will die, and such inevitability must be respected and known, yet not cause dread. No more fear taxes, since he who earns can scarce keep it all, can he, any more than he can take it with him? For what purpose would you keep it all? I grant you death and taxes should be avoided - where it's possible to do so with honesty and honour. But they are not worthy of our fear.

Maybe boredom is a more suitable object. To live an uninteresting life is both eminently possible and profoundly fearful. Don't you have a mind though, and doesn't it engender imagination? And is it beyond that imagination to find an action, project or enterprise that could engage you in some means of improving your own or your fellow's lot? Boredom is not worthy of fear, since it is easily remedied.

Yet I think there is an entity worthy of fear, fear true and cold. It is waste... the waste of human potential. Such waste inhabits every corner of life where ideas spring up and are swallowed, spontaneously springing to life and dying like fundamental particles separating and colliding undetectably in the depths of space. Waste of human potential is omnipresent. It occurs in places of learning, and places of work, in families and in churches, in friendships and in war. It steals not only from those whose potential is untapped but from their communities, their
 relationships and indeed all society.


Fear this waste. Yes. Fear it and fight it.

Thursday, 11 July 2013

More Musings on Little's Law

My previous posting about why not  to use Cycle Time in Kanban resulted in some interesting discussions, and I'm grateful to +Steve Tendon for pointing me in the direction of this paper [1] by John D.C. Little and Stephen C. Graves which gives some very helpful historical background into the derivation of Little's Law, its applicability and some of the terminology used.

From Little and Graves (2008)
Little's own formulation of the "law" was as follows:

L=λW

where

L = average number of items in the queuing system,
(equivalent to WIP in Kanban terminology)

W = average waiting time in the system for an item,
(equivalent to System Lead Time)

λ = average number of items arriving per unit time
(equivalent to Delivery Rate, assuming "stationarity")

With Kanban preferred terms we can see this maps to:

WIP =  Delivery Rate * Lead Time
or
Delivery Rate = WIP / Lead Time

Little used "waiting time" for the time taken by one unit to traverse the system (W) because his original context was queuing systems. For other applications he suggested Flow Time, which I think is a very useful alternative.

He also notes though that other authors use other terms for W, including cycle time, throughput time, and sojourn time, depending on the context. Yes - cycle time I'm afraid is in that list which is why confusion still abounds. This conflicts with the more generally accepted definition of cycle time in manufacturing, which corresponds to the target rate of working expressed as Takt Time, and is the reciprocal of Delivery Rate. In other words this confusion of terminology is at least as old as the reference Little and Graves cite: Factory Physics by Hopp and Spearman (1st edition:1996).

Useful background, but the message to me is still: "Don't use Cycle Time in Kanban!".


References:

[1] Little, J. D. C and S. C. Graves (2008). Little's Law, pp 81-100, in D. Chhajed and TJ. Lowe (eds.) Building Intuition: Insights From Basic Operations Management Models and Principles. doi: 10.1007/978-0-387 -73699-0, (c) Springer Science + Business Media, LLC http://web.mit.edu/sgraves/www/papers/Little%27s%20Law-Published.pdf

[2] Hopp, W. J. and M. L. Spearman (2000). Factory Physics: Foundations of Manufacturing Management, 2nd (ed.), Irwin McGraw Hill, New York, NY.

Thursday, 4 July 2013

The difference between Cycle Time and Lead Time... and why not to use Cycle Time in Kanban

Before addressing the question of which terms you should use in the Kanban method, let me attempt an explanation of the generally accepted meaning of these terms. (We'll come to the question of how generally later!)

Firstly Cycle Time: it is the time between units emerging from a process. You could visualise it like this:


Our system here could be a Development Team say, and the units User Stories (although we would expect to have more variation in this case). Equally it could be a bicycle assembly plant (working not very fast). The Cycle Time here is half a day because the system produces one unit every 0.5 days. Therefore the Delivery Rate (which is Kanban's preferred term for measuring throughput) is 2 units per day.

To understand Lead Time we need to put a marker on one of the items entering the process, and then see how long before that particular item emerges. Like this:


Lead Time is the time taken for one unit to pass through the process. So our Lead Time here is 5 days.

Now before we consider Little's Law, can you look at those 2 pictures and tell me how many units of work there are in progress in the system under study (WIP)?

Scroll down when you've thought about it...






Scrolling...






Still Scrolling...






Nearly there...




Anyone say 10? That wasn't too hard was it? You've just derived a simple application of Little's Law for a regular (and "stationary") process, and you probably found it easier than plugging the numbers into Little's Law itself:

Delivery Rate = WIP / Lead Time

or

WIP = Delivery Rate * Lead Time
WIP = 2 * 5 = 10

So Cycle Time (as defined here) and Lead Time are very different concepts. However there is one important case where they are equal... when there is only one unit in progress in our system (plug WIP=1 into the formula if you doubt this, or change the Lead Time in the example above to 0.5 days and visualise what is happening). Whenever WIP>1, Lead Time will be longer than Cycle Time.

Why not use Cycle Time in Kanban?

Ok so if that's what the terms mean, why shouldn't we use Cycle Time in a Kanban context?

Put simply, because people use different terms to mean the same thing, and - even worse - the same term to mean different things. And the worst confusion is associated with the terms Cycle Time and Lead Time.

There are authoritative sources from well-respected authors that use Cycle Time to mean something different to its common usage in manufacturing circles (the meaning explained above). This inevitably leads to a whole bunch of confusion. This is why a session at the recent "Kanban Leadership Retreat" (#klrat) led by +Dan Brown sought a common policy for use of these terms in Kanban. They proposed a first step which was that the terms recommended by Kanban for use in Little's Law should be Delivery RateLead Time and Work In Progress (WIP). Furthermore it proposed that the use of Cycle Time should be deprecated, and its reciprocal Delivery Rate should be used instead.

I happen to agree and support this proposal, but I also think there is a necessary second step: to understand what Cycle Time is, as defined in manufacturing, so that you can explain why we use Delivery Rate in Kanban instead of this term as the measure of throughput, and also so that you can explain the Kanban terminology to those who wish to use Cycle Time and Lead Time as synonyms... and hopefully persuade them to adopt the agreed Kanban terms instead.

Good luck!

See also: Why I don't use cycle time in Kanban and Glossary Proposal

Notes:

See also: the  list of terms and definitions that you need when measuring flow and efficiency in Kanban, posted previously.

I'm indebted to this Harvard Business School archive from the year 2000 (which uses Professor W. Bruce Chew's glossary) for the clearest explanation of these terms I've seen so far [http://hbswk.hbs.edu/archive/1460.html], also for the discussion of cycle time In Womack and Jones' "Lean Thinking" (1996, 2003),  and for this blog [http://www.isixsigma.com/methodology/lead-time-vs-cycle-time/] from Fang Zhou which also gives a very clear explanation and was my source for the HBS page .

I'm also very grateful to all those from the Kanban community who have put up with my badgering  and whining about this issue over the last few weeks! Their comments have helped me clarify the issue and decide how best I can explain it to my clients.

In accepting Lead Time as the preferred term for the time taken by one unit passing through the system, the Kanban community has to acknowledge some problems, particularly that some people wish to use this term only for the time from customer order to delivery (which is why qualifying this term, e.g. System Lead Time, may be needed). Little himself used "Wait Time" (because his context was queuing systems). However he also suggested the term Flow Time, which is certainly another good candidate for an unambiguous term.

Wednesday, 26 June 2013

There are no flavours of Scrum

Here's a great posting from Jem D'jelal on the trials and tribulations of a London Scrum Master "Since When Did Ben & Jerry's Create Scrum?" I smiled throughout and recognised some of the frustrating situations that I also hit as a Scrum Master. However I realised Jem was also eloquently expounding some of my uneasiness with Scrum. Here's the stand-out section for me:

Whoever you are, please read and repeat the following mantra after me:
There are no flavours of Scrum. 
There is no chocolate flavoured Scrum.
There is no Caramel chew chew flavour of Scrum.
Scrum is not an ice cream.
It does not come in bloody flavours.
If you want to pick an Agile framework, methodology, technique, or approach which comes in different flavours, go for one which is less prescriptive than Scrum. That is completely fine. Kanban for one example (There are only 2 principles, it leaves almost everything open. The only rules are you need to visualize your workflow and limit your WIP).
But Scrum does have rules... if you do not follow these 9 rules then you are not doing a different ‘flavour’ of Scrum – you are simply NOT doing Scrum.
But doesn't this make you uneasy Jem? Don’t you worry (like me) that as agilists we want to embrace change while at the same time saying these particular practices are eternal? Many changes from a good method like Scrum are bad (measured say by a reduction in the flow of value in this context). But does that imply every conceivable change, even from a process that is working exceptionally well, will always be bad?

And where does the idea come from that Kanban has two principles (see my post "There are 3, no make that 4... no really, shouldn't it be 3 Principles of Kanban?"). The two rules (i.e. practices that if you are not doing, you are definitely not doing Kanban) come from Jim Benson's Personal Kanban book which is just the starting point of exploring this method. Don't be fooled this sums up the whole approach. (These "rules" are the first two of the six practices of Kanban.)

Since none of the principles of Kanban (whether there are 2, 3 or 4) mention WIP or visualisation shouldn't we take a look at the principle that is relevant here, and the one that makes me comfortable to be a Scrum Master using Scrum with all its rules intact (except the one about the rules being immutable).

The principle is “agree to pursue evolutionary change”. This implies you’ll try some changes that don’t work and don’t get amplified. But also if some changes work, you’ll amplify them even if they break the previous policy you were applying.

We use this in the Retrospective (which incidentally we still do every Sprint) to encourage each other to find small changes - for example to the Definition of Done (DoD) - that might improve the situation we're in. For instance if there are too many bugs coming back at us from the Release process or the field, we ask whether a tighter DoD might help. If the change works we keep it and tell other people about it (select, amplify). If we can't tell whether it's better, we'll probably keep it (diversity in the ecosystem). If it makes things worse we regress (deselect). In other words we're using an evolutionary change mechanism.

Our Scrum team limits work in progress by not putting too much into the Sprint Backlog and by focusing on finishing stories we've started rather than starting new ones. And we visualise most of our work using a Scrum board and burndown charts. Why should anyone tell me I am not using both Scrum and Kanban just because there's one Scrum rule that I refuse to accept in my context... that the rules are immutable?

Now our use of Kanban is not very deep at this point and there are several practices where we're certainly deficient. (Not following all the Kanban practices doesn't mean we are not doing Kanban by the way, which is a key difference from Scrum). Take Kanban practice number 3:"Manage Flow". We've begun to gather metrics on elapsed time for stories from entering In Progress to Done (Development Lead Time), but we haven't gone wider yet in the value flow, nor used this to ensure smoother and more flow-efficient ways all the way from "bright idea" to released, which is our ambition. There should be some good improvements we can get here in the future.

I admit after some period of time of changing our process, you may tell me Jem that I'm not even using a flavour of Scrum, it's just not Scrum. Probably at that point I won't mind too much (unless of course you were to interview me and not give me a job as a result!).

Thursday, 13 June 2013

What is Scrumban?

What is "Scrumban"? The name seems to offer a simple answer - surely it's a mixture of Scrum and Kanban? So you think the rules of Scrum are a bit strict. You think Kanban doesn't offer enough guidance on things like roles and when planning and retrospectives should take place? Why not just mix the two and things are bound to be better!

The only snag with this is that Scrum and Kanban already have definitions which are not really compatible with this mixture idea. And Scrumban, at least in the original usage of the term, was coined with a different meaning than mixing the two methods (see Scrumban, Corey Ladas 2009).

The definition of what is and isn’t Scrum is well understood, thanks to the concise publication of its “rules” in the Scrum Guide. Scrum is a process framework – a set of guidelines and constraints within which your team can define and improve the process you use for developing products. It’s not a process as such - it is silent on many of the aspects a process needs (it says you need a “Definition of Done” for example, but not what it should be). However it is a framework that is very specific about some aspects of the process, such as when planning meetings take place and how long they should last.

Scrum is often contrasted with Kanban, also a framework, but more accurately described as an improvement framework; that is the approach for assessing and improving your process. Kanban has fewer constraints or rules than Scrum, allowing a wide range of processes from those based on Scrum itself to many variants – agile and less agile – which can be the process's starting point . As +David Anderson explains, "A Kanban system is overlaid on an existing process." Provided you can look at your work as a flow of items - usually this is simply a case of changing the way you look at your work - you can use Kanban to assess and improve your process.

As yet there is not the simple definition booklet that defines what is and is not Kanban, so the best starting point is the “Blue Book”, Kanban (Anderson, 2007), plus the summaries - published informally since that time - of the foundational principles and core practices. Because I like short definitions, I've summarised these as "how to adopt Kanban":
  1. See work as flow (Lean Flow Paradigm)
  2. Start from here and improve (Foundational Principles)
  3. Make work and policies explicit;
    Make validated improvements (Core Practices)
If you are doing this, you are doing Kanban, even if your policy for work in progress (usually cited as a key difference between Scrum and Kanban), is to limit work in a Sprint to that which the team believes it can complete (the policy that Scrum applies)... or if you decide not to limit work in progress at all.

In fact there is nothing in Scrum that is incompatible with adopting Kanban as well. Nothing that is except the “rule” that the rules can’t change. If you can validate that it would be an improvement to do planning more or less frequently, or to decouple the frenquency at which you do retrospectives from the frequency at which you do reviews, Kanban would say you should make that change. Scrum might say okay make the change too, but if you do so you are no longer doing Scrum.

So this is the key to what Scrumban is. Scrumban is a process being improved using Kanban, which probably is no longer strictly Scrum. Some part of the strict Scrum framework has been modified (or maybe was never tried) because it is believed the change is more appropriate, brings more benefits or incurs less cost than pure Scrum. Such processes used to be referred to, somewhat derogatorily as “Scrumbut”. Scrumban is the more acceptable, less pejorative alternative... and it's also more positive because it implies you're using Kanban.

In his book on Scrumban, Corey Ladas describes a possible trajectory for a process that starts off as Scrum and changes as Kanban improvements are applied. This is what he meant by “Scrumban”. At the end of the story the process has very little similarity to Scrum and few people would even describe it as “Scrumban”. It is simply a flow-based process, as near as possible to single piece flow, with a typical Kanban improvement process around it. Is this process still Scrumban? At the point when the similarities to Scrum have disappeared, I’d suggest the label is not helpful. Many have concluded the label itself is not helpful anywhere since it causes the kind of confusion that is visible in statements like "I prefer Scumban to Kanban" or "Kanban is a better process than Scrum". Look at the definition of these terms and it’s clear such statements are not really meaningful.

I think though that a commonly accepted definition of Scrumban would be useful, at least to reduce the confusion in discussions around this topic. So here’s my attempt:

“Scrumban is a Scrum or Scrum-like process which is being improved with Kanban.”

This definition implies that if you are doing Scrumban you are using Kanban! Maybe that will be difficult to accept for some people, particularly if they believe their approach is "better than Kanban". Maybe a Scrumban community will emerge and wish to define distinctive practices from Kanban, just as most Scrumban processes differ from Scrum in some (albeit different) ways. In the meantime I’ll continue to use the term Scrumban - guilty secret: yes I do use the word! And when I do, that’s what I mean by it.

Postscript:
You can't do Kanban in a vacuum. You need a starting point. If your starting process is based on Scrum or is Scrum-like, then according to this definition it is Scrumban. If the starting point was Prince II or XP or DSDM should you call it Princeban, XPban or DSDMban?!